Analog Circuit Fault Diagnosis Based On Hybrid Optimization Neural Network

Currently,analogue electronic circuits are widely used in electronic systems in aerospace,military,communication and other areas.Statistics show that in most hybrid circuits,compared to digital circuits,the failure rate of analog circuits is higher,which reduces the reliability of the products,and in most electronic systems almost 80% of errors occur in analog circuits.However,the current analog circuit error diagnosis methods are inefficient and do not adapt to the rapid development of analog circuits.Therefore,it is necessary to study the fault diagnosis technology for analog circuits.Firstly,aiming at the difficulty of nonlinear feature extraction and feature extraction of progressive soft fault output response signal of analog circuit,and the problem that the signal can’t be diagnosed and classified accurately,an immune genetic algorithm is proposed to optimize the parameter optimization process of BP neural network,so as to realize analog circuit fault diagnosis.Using wavelet packet analysis,the output response of the analog circuit is decomposed and reconstructed by four-layer wavelet to complete the extraction of feature vectors;then immune genetic algorithm is used to optimize BP neural network for training and testing to achieve fault diagnosis for different faults.Secondly,there are some problems,such as the tolerance of analog circuit leads to the decline of diagnosis effect.A fault diagnosis method based on FCM and genetic algorithm is proposed to optimize probabilistic neural network.The FCM is used to fuzzy cluster a group of feature vectors,and the pre-velocity of each feature vector is calculated,and then the feature vector which meets the requirement of pre-speed is obtained.The genetic algorithm is used to solve the main parameters in the probabilistic neural network.The optimal decomposition of smoothing parameters is obtained to achieve the purpose of optimizing the diagnosis effect.Finally,the simulation experiments are carried out by using the basic partial voltage amplifier circuit and the fourth operational amplifier second-order filter circuit respectively to verify the above methods and carry out comparative experiments.The fault data are obtained by Multisim simulation circuit,the fault data are extracted by MATLAB software,the feature vector set is obtained,and diagnostic experiments were carried out.The results show that the proposed method can greatly improve the accuracy and accelerate the training and testing speed of the algorithm at the same time.